Skip to main content

AI-enabled grievance & ticketing

Facility management,finally made wise.

From “the AC isn’t working” to “resolved, with proof” — TicketWise triages every complaint with AI, routes it to the right crew, and closes the loop on SLA. Built for India’s hotels, hospitals, corporates and malls.

Book a demo See the workflow
  • Triage at the speed of typing
  • Proof, not promises
  • Your facility runs 24×7 — so does TicketWise
  • Every ticket, handled wise
Live queue · indicative
T-2847RESOLVED

Water leak under sink · 3F restroom

HVAC · 1h 14m

T-2851TRIAGED

Spillage in 2F corridor

Housekeeping · auto-routed

T-2860NEW

AHU rattling · 4F server room

Just raised

TicketWise by the numbers

1000+
Tickets handled / month

across pilot clients

94%
Resolved on first pass

verified by photo proof

70%
Auto-routed, zero touch

deterministic round-robin

<2h
Median resolution

for standard severity

The workflow

Watch a ticket move through TicketWise

One complaint, seven steps — raised, triaged by AI, routed, dispatched, proven, resolved and closed. On a wide screen, scroll to ride along.

  1. NEW

    Step 1 · Origin

    A complaint is raised

    A client reports a facility issue — a leak, a spill, a CCTV fault. The ticket carries title, photos, location and service category, scoped to their organisation from the first keystroke.

    Org-scoped intake

  2. TRIAGED

    Step 2 · AI Layer

    AI triages it in seconds

    An LLM predicts service category and severity within seconds, then hands the supervisor a banner to accept or override. Every call is logged, replayable, and falls back to a smaller model on failure.

    Llama 3.3 70B · classify_ticket

  3. ASSIGNED

    Step 3 · Assignment

    Routed to the right hands

    About 70% follow deterministic round-robin to the next available team lead — no human in the loop. The rest (high severity, low confidence, escalations, reopens) land on the supervisor queue.

    Auto round-robin · supervisor fallback

  4. DISPATCHED

    Step 4 · Service Teams

    Dispatched to the crew

    Twelve categories collapse into three operating lines — Housekeeping & Pantry, Security & Facade, and HVAC / Electrical / Plumbing. The team lead nominates a specific field worker from the on-shift roster.

    3 operating lines

  5. PROOF

    Step 5 · Field Execution

    Work done, proof captured

    The worker completes the job and uploads 1–3 photos plus notes. A vision model auto-captions the evidence and flags mismatched scenes, so a supervisor can verify without opening every image.

    Vision AI · analyze_image

  6. RESOLVED

    Step 6 · Resolution

    Stop the clock

    Proof accepted, the SLA clock stops. If the job can’t be completed — missing part, unsafe access, out of scope — it escalates back to the supervisor with the original SLA preserved.

    SLA met · or escalate

  7. CLOSED

    Step 7 · Final State

    Closed & logged

    The client confirms (or the reopen window lapses) and the ticket closes. A weekly AI digest quotes one-line outcomes to the COO, grouped by client and service line. Reopens loop straight back to the queue.

    Weekly COO digest

Escalations and reopens loop straight back to the supervisor queue — with the original SLA preserved, so nothing slips through the cracks.

Capabilities

Everything a facilities desk needs — in one calm queue

From a leaking guest room to a stalled mall escalator, every grievance flows through the same calm pipeline — triaged, routed, proven and closed on SLA. No spreadsheets, no WhatsApp threads, no guesswork.

  • AI

    AI triage in seconds

    Llama 3.3 70B predicts category and severity the moment a ticket lands, with a smaller model as automatic fallback. Supervisors accept or override — never start from a blank form.

  • Routing

    Smart auto-routing

    Deterministic round-robin sends ~70% of tickets straight to the right team lead. Only the genuinely tricky ones reach a human — high severity, low confidence, escalations and reopens.

  • Proof

    Photo-proof verification

    Field workers upload photos and notes; a vision model auto-captions them and flags scene mismatches. Supervisors verify in a glance — proof, not promises.

  • SLA

    Live SLA clocks

    Every ticket shows response and resolution countdowns — green, amber, breached — carrying the time as text, not colour alone, so it works without colour vision.

  • Security

    Bank-grade multi-tenancy

    Every record is scoped to an organisation at the query layer. Cross-tenant access returns a 404, never a 403 — so one client can never even learn another exists.

  • Loops

    Feedback loops that hold

    Escalations and reopens route back to the supervisor automatically — and SLA invariants are preserved, so a reopen can’t game a first-response promise that was already kept.

  • Alerts

    Instant Telegram alerts

    High and critical severity pings the supervisor channel the moment it happens — escalations, reopens and breaches surface before a client has to chase.

  • Reporting

    Weekly COO digest

    An AI summary condenses long ticket histories into quotable, one-line outcomes — grouped by client and service line, delivered every Monday at 08:00 IST.

  • Observability

    Full AI audit trail

    Every model call logs the model used, latency, tokens, prompt and response — replayable and measurable, so you can see exactly where AI is earning its keep.

Proof, not promises

Every closure is backed by evidence

A worker can’t just mark a job done. They capture it — photos, notes, a timestamp — and a vision model checks the scene before a supervisor signs off.

  • Audit-ready by default

    Every closure carries a timestamped photo and note — the kind of trail a hotel or hospital audit can replay months later.

  • Fewer reopens

    A vision model checks the scene before sign-off, so jobs marked “done” actually are — and clients stop reopening them.

  • Client trust, earned

    Corporate parks and malls see what was fixed, not just a status flip. Proof turns “trust us” into “see for yourself”.

Verified

14:32 IST · 1h 14m

Ramesh Yadav

HVAC field technician

Ticket
T-2847 · Water leak under sink, 3F restroom
Client
5-star hotel · Aerocity, New Delhi
Before
Visible leak at PVC joint
After
Joint replaced, dry after test run
AI

AI caption: “Replaced PVC joint under sink, no leak observed after 5-min flow test.”

Replaced 32 mm PVC elbow joint and re-sealed with thread tape. Ran 5-min flow test, no seepage. Cleaned cabinet interior.

The AI layer

AI that earns its keep — and proves it

Every model call funnels through one service. There is one place to swap a model, one place to add observability, and one place to enforce rate limits. Nothing about the AI is a black box.

  • One shadow service

    Pick the model by task, retry with exponential backoff, fall back to a smaller model, and persist the full prompt and response to an audit table.

  • Model routing by config

    Model choice is a deploy-time decision, not a runtime one. Swapping a model is a config change — code reads the route by name.

  • Per-org rate limits

    A token bucket (30 RPM, configurable per org) keeps a misbehaving prompt loop from ever blowing the budget.

  • Versioned, replayable prompts

    Prompts are versioned so prior runs replay deterministically — the substrate for prompt iteration and regression testing.

ai_serviceMODEL_ROUTES
Live
  • classify_ticket
    Llama 3.3 70BPredict category & severity
  • analyze_image
    Llama 3.2 90B VisionCaption & QA proof photos
  • summarize_ticket
    Llama 3.3 70BCondense history for the digest
  • assist_field
    Llama 3.1 8BExpand thin descriptions
  • draft_reply
    Llama 3.3 70BSuggest client responses
  • nl_to_filter
    Llama 3.1 8BPlain-English ticket search
every call logged · latency · tokens · prompt & response

Industries

Built for India’s facility teams

From five-star hotels and hospitals to corporate parks, malls and sprawling campuses — if your team fields a steady stream of maintenance and housekeeping requests, TicketWise keeps every grievance triaged, routed and resolved with proof.

  • 01

    Hotels & Hospitality

    Guest rooms, banquets and back-of-house — where a slow fix is a bad review. Round-robin keeps housekeeping moving without a supervisor in the loop.

  • 02

    Hospitals & Healthcare

    Hygiene-critical, fast-SLA environments. Photo proof and tight escalation paths keep compliance auditable.

  • 03

    Corporate Parks & Offices

    Multi-tower campuses and managed offices. Strict org scoping keeps every tenant’s tickets fully isolated.

  • 04

    Malls & Retail

    Footfall-heavy facades, HVAC and escalators. Vision AI captions evidence so the right crew is dispatched the first time.

  • 05

    Education & Campuses

    Hostels, labs and grounds across sprawling campuses. One queue, one audit trail, many buildings.

  • 06

    Co-working & Managed Spaces

    Many brands under one roof. Per-org rate limits and tenant isolation keep operations clean and billable.

FAQ

Questions, answered

  • What is TicketWise?

    TicketWise is an AI-enabled grievance and ticketing platform for facility management teams. It triages complaints with AI, routes them to the right crew, captures photo proof of the work, and tracks SLAs from raise to closure.

  • Which teams is it built for?

    Facility management operators serving hotels, hospitals, corporate parks, malls, education campuses and co-working spaces — anywhere a team handles a steady stream of maintenance and housekeeping requests.

  • How does the AI triage work?

    When a ticket is raised, a large language model predicts its service category and severity within seconds. The supervisor sees the suggestion as a banner they can accept or override. Every call is logged and falls back to a smaller model if the primary is unavailable.

  • Is my data secure and isolated?

    Yes. Every record is scoped to an organisation at the query layer, and cross-tenant access returns a 404. TicketWise is built to be compliant with India’s Digital Personal Data Protection Act, 2023, with a named Grievance Officer for data-principal requests.

  • Can supervisors override the AI?

    Always. The AI suggests; humans decide. Supervisors can accept, override or re-prompt the classification, and they can manually route any ticket.

  • What happens when a worker can’t finish a job?

    They escalate from the field with a reason and a photo. The ticket returns to the supervisor queue with its original SLA preserved, and high-severity escalations fire an instant alert.

Ready when you are

Bring order to the chaos.

Every complaint, closed before it becomes a crisis.

Book a demo Email us